Turnitin (TII) has developed a new tool for checking for potential artificial intelligence (AI) generated text in submitted assessments, which will go live for UNE from Wednesday 5 April 2023. Here’s some key initial information about the tool, how it will be implemented and some initial guidance on its use.
What is the AI detection tool?
The tool is an AI model developed by TII that is trained on a range of AI and human-generated text, and analyses text based on word sequence probability.
The tool runs in the same way as the similarity checker when a file is submitted, and the result is displayed as an AI indicator in the similarity report. When the similarity report is accessed, an AI score will be displayed separately to the similarity score. Clicking through the indicator will display the AI score, with highlighted text flagged by the tool and links to resources. The AI indicator score is displayed to admins and instructors only and will not be available to students.
It is worth noting the current, identified limitations of the tool. The AI score is simply an attempt by the TII model to identify text that may have been generated by AI based on a single parameter. The TII model is also trained on GPT-3 and GPT-3.5 language models, which are not the most recent versions.
Additional information from TII on the tool is available in the Teaching Online @UNE site.
How will it be implemented?
An additional flag with the AI score (see example below) will be included in the TII similarity report from Wednesday 5 April 2023.
Initial guidance for use
The tool should be used in the same way as the current similarity checker – that is, it may detect concerns around academic integrity in the submitted work that should be considered further. We recommend that you continue to consider all aspects of the submitted work in considering whether there is cause for concern (TEQSA’s guide for detecting contract cheating may assist with aspects to consider).
If following this consideration, there is still cause for concern around the unauthorised, substantive use of AI-generative tools in submitted work, the assessment can be referred to your School’s Academic Integrity Investigating Officer for investigation.
We note that the tool has been developed at speed to respond to recent developments in AI-generative tools. Consequently, as with all new developments, we will learn more about the effectiveness and limitations of the tool as it is used here and across the sector. Education Futures will be working with the Academic Integrity Investigating Officers to collect data from UNE and the sector, monitor use and trends, and develop more comprehensive guidance and processes as the understanding of the tool develops. Any updated guidance will be provided via this blog, your School’s Academic Integrity Investigating Officer and Teaching Online @UNE.